Close ✕
Science and tech U.S.

Google’s Smart Glasses Put Privacy to the Test

A pair of camera-equipped smart glasses displayed against a plain background.
Photo: Erick Miller / Wikimedia Commons, CC BY-SA 2.0

Google smart glasses arrive with a familiar promise: make artificial intelligence easier to reach by putting a camera, microphones and a Gemini connection on your face. They also arrive with a familiar problem. People nearby may not know when they are being recorded, and an indicator light does not settle the question of consent.

The glasses are a joint project by Google, Samsung and eyewear brands Warby Parker and Gentle Monster. Samsung supplies the hardware; Google supplies the software. Buyers can preorder the first models now, and sales are scheduled to begin December 10. Warby Parker’s range starts at $395 and Gentle Monster’s at $449, before any added cost for prescription lenses. The frames have no displays in their lenses. Users can speak to Gemini or tap the frame, and can use the camera to ask about what they see. The report on the new Android XR eyewear (R4) covers the models and launch details, while the product overview gives more on what the glasses can do.

Google is trying to distinguish its offering from Meta’s popular camera glasses through a layered privacy pitch. A recording light comes on when a photo or video is captured; a physical switch can turn the glasses off. Google says the camera stops working if someone covers or alters its light. Google limits recording clips to 90 seconds, and users have to agree to avoid recording people without their consent, filming in private or sensitive settings, and interfering with the indicator. Violating those terms could cost a user access to the glasses. Setup also requires an account holder to be at least 18.

The more unusual protections concern what happens when Gemini looks through the camera. Google says it redacts background faces before the image reaches Gemini, and does not save visual data from such inquiries in Gemini History, use it for advertising or use it to train its models. Google encrypts the data on the device and in transit if it must send it elsewhere for processing. Directional microphones aim to isolate the wearer’s voice, with a further noise-reduction step to keep nearby conversations from being sent to the AI. Camera access by third-party apps requires permission, which users can revoke. App-store rules are intended to bar real-time facial recognition, social livestreaming and attempts to change the recording light’s behavior. Google’s privacy plan and a closer look at the glasses’ privacy limits describe these controls.

There is an important boundary to those assurances. Face redaction applies when a user asks Gemini a question through the camera; it does not mean Google automatically removes faces from ordinary photos or videos a user chooses to save. Google’s policy also cannot make a bystander aware of the recording light, or make that person comfortable with it. In a separate hands-on demonstration, Gemini translated conversations across multiple languages and handled noisy scenes better than the tester expected. It also refused requests to identify people in photos or video. That refusal reflects a deliberate choice to keep face identification out of the feature, not simply a promise to use it carefully. The hands-on account of those tests is here.

Google says its obligations extend beyond the glasses themselves. It plans to prohibit sales and ads for accessories made to conceal recording lights. YouTube’s rules prohibit harassment and unwanted sexualization, including in footage recorded on glasses, though those restrictions apply to video from phones and other cameras too. Google Play rules would also restrict certain third-party uses of the camera and microphone, subject to review. The company’s case is that a platform owner can set rules across hardware, software and services. Meta does not control an equivalent app store or computing platform, a difference that limits direct comparisons, though it does not erase the contrast in the policies described.

The unresolved issue is the social setting. Google is researching ways to alert people when the glasses are nearby and has considered location-based restrictions. Neither idea is ready as a solution. People may circumvent location controls by disconnecting a phone, and the controls would raise questions about the use of location data. Phone alerts could become noise, and an alert alone may leave a person unsure what to do. A bystander may not feel able to challenge a stranger, or someone with authority at work. Privacy specialists have also noted that some people may find an indicator harder to interpret: a person with limited sight may not see it, and others may not understand what it signals. The broader drawbacks of AI glasses are not merely technical.

The market is growing despite the unease. Shipments climbed 54 percent from the previous year in the second quarter of 2026, and one eyewear company said seven million pairs of its AI glasses sold in 2025. Those figures help explain why Google and Meta keep pressing ahead. They do not establish public acceptance. Google’s new protections may reduce some risks, especially when Gemini processes visual information, but the company has not shown that bystanders will regard a small light and a set of platform rules as enough. Research into notifications and location-based limits remains early, and the glasses’ launch will put the promises under public scrutiny.

For my part, I find the face-redaction choice more meaningful than another assurance printed in a policy. But the person across the table still has to live with the camera. That is the test the launch cannot answer in advance.

Reported by The Verge, Wired, CNN, Ars Technica.

How the story went

Related stories

Top stories now

About